Add support for undocumented tasks

[Kardbord]

The following tasks are available in the model hub, and seem to have inference support, but are not yet listed in the Inference API docs.

• [x] Text To Image
• [x] Image to Text
• [ ] Visual Question Answering
• [ ] Document Question Answering
• [ ] Zero Shot Image Classification
• [ ] Text to Speech
• [ ] Audio to Audio
• [ ] Tabular Classification

[Kardbord]

Text to image support added in #16

[Kardbord]

It seems I was looking in the wrong place for diffuser docs. They can actually be found here: https://huggingface.co/docs/diffusers/using-diffusers/pipeline_overview.

Also see https://huggingface.co/docs/huggingface.js/inference/README. Many tasks that are not documented in the official inference docs are implemented there.

• [x] Add links to diffuser and huggingface.js docs in README

[Kardbord]

Image to text support added in #19

[Kardbord]

I stumbled across a JSON schema for the Interence API which seems to be more up to date than the docs page on the HF website. This will be handy for both implementing undocumented tasks and maintaining existing ones.

https://huggingface.github.io/text-generation-inference/openapi.json